epidemiology concept of cause

8/3/2019 Epidemiology Concept of Cause

1/45

Epidemiology

Concept of Cause


2/45

A major focus of epidemiology is informing

efforts to prevent and control disease and

promote health. To do this, we need to knowthe causes of disease or injury and the ways

in which these causes can be modified. This

chapter describes the epidemiologicalapproach to causation.


3/45

The Concept of Cause An understanding of the causes of disease or

injury is important not only for prevention, butalso for correct diagnosis and treatment. Theconcept of cause is the source of muchcontroversy in epidemiology. The process bywhich we make causal inferences judgmentslinking postulated causes and their outcomes is a major theme of the general philosophy ofscience, and the concept of cause has differentmeanings in different contexts.


4/45

Sufficient or necessary

A cause of a disease or injury is an event,

condition, characteristic or a combination of

these factors which plays an important role inproducing the health outcome. Logically, a

cause must precede an outcome. A cause is

termed sufficient when it inevitably produces or

initiates an outcome and is termed necessary ifan outcome cannot develop in its absence.


5/45

Multiple Factors

A sufficient cause is not usually a single factor,

but often comprises several components (multi-

factorial causation). In general, it is notnecessary to identify all the components of a

sufficient cause before effective prevention can

take place, since the removal of one component

may interfere with the action of the others andthus prevent the disease or injury.


6/45

Attributable fraction

The attributable fraction can be used to quantifythe likely preventive impact of eliminating aspecific causal factor.

The attributable fraction (exposed), also knownas the etiological fraction (exposed), is theproportion of all cases that can be attributed to aparticular exposure. We can determine the

attributable fraction (A

F) by dividing the risk (orattributable) difference by the incidence amongthe exposed population.


7/45


8/45

Causal Pathway Epidemiologists have been criticized, particularly by

laboratory scientists, for not using the concept of causein the sense of being the sole requirement for the

production of disease. Such a restrictive view ofcausation, does not take into account the fact thatdiseases commonly have multiple causes. Preventionstrategies often need to be directed simultaneously atmore than one factor. In addition, causes can be linkedto a causal pathway where one factor leads to another

until eventually the specific pathogenic agent becomespresent in the organ that gets damaged; this can also becalled a hierarchy of causes.


9/45


10/45

It is often possible to make major progress

in prevention by dealing only with the more

remote or upstream causes. It was

possible to prevent cholera cases decades

before the responsible organism let

alone its mechanism of action had beenidentified (Figure 5.3).


11/45


12/45

Single and

Multiple Causes

Pasteurs work on microorganisms led to the formulation,first by Henle and then by Koch, of the following rules fordetermining whether a specific living organism causes a

particular disease: The organism must be present in every case of the

disease;

The organism must be able to be isolated and grown inpure culture;

The organism must, when inoculated into a susceptibleanimal, cause the specific disease;

The organism must then be recovered from the animaland identified.


13/45

Anthrax was the first disease demonstrated to meetthese rules, which have since proved useful with manyother infectious diseases and with chemical poisoning.

However, for many diseases, both communicable andnon-communicable, Kochs rules for determiningcausation are inadequate. Many causes act together,

and a single factor such as tobacco use may be acause of many diseases. In addition, the causativeorganism may disappear when a disease hasdeveloped, making it impossible to demonstrate theorganism in the sick person. Kochs postulates are of

most value when the specific cause is a highlypathogenic infectious agent, chemical poison or otherspecific factor, and there are no healthy carriers of thepathogen: a relatively uncommon occurrence.


14/45

Factors in CausationFour types of factors play a part in the causation of disease, all may benecessary but they are rarely sufficient to cause a particular disease orstate:

Predisposing factors, such as age, sex, or specific genetic traits thatmay result in a poorly functioning immune system or slow metabolism of

a toxic chemical. Previous illness may also create a state ofsusceptibility to a disease agent.

Enabling (or disabling) factors such as low income, poor nutrition, badhousing and inadequate medical care may favour the development ofdisease. Conversely, circumstances that assist in recovery from illnessor in the maintenance of good health could also be called enablingfactors. The social and economic determinants of health are just as

important as the precipitating factors in designing preventionapproaches.

Precipitating factors such as exposure to a specific disease agent maybe associated with the onset of a disease.

Reinforcing factors such as repeated exposure, environmentalconditions and unduly hard work may aggravate an established disease

or injury.


15/45

The term risk factor is commonly used todescribe factors that are positively associatedwith the risk of development of a disease butthat are not sufficient to cause the disease. The

concept has proved useful in several practicalprevention programmes. Some risk factors (suchas tobacco smoking) are associated with severaldiseases, and some diseases (such as coronaryheart disease) are associated with several riskfactors (Figure 5.4).


16/45


17/45

Epidemiological studies can measure therelative contribution of each factor to disease

occurrence, and the corresponding potential

reduction in disease from the elimination of each

risk factor. However, multi-causality means that

the sum of the attributable fractions for each risk

factor may be greater than 100%.


18/45

Interaction The effect of two or more causes acting together

is often greater than would be expected on thebasis of summing the individual effects. Thisphenomenon, called interaction, is illustrated bythe particularly high risk of lung cancer in peoplewho both smoke and are exposed to asbestosdust (Table 1.2). The risk of lung cancer in thisgroup is much higher than would be indicated bya simple addition of the risks from smoking (tentimes) and exposure to asbestos dust (fivetimes); the risk is multiplied fifty times.


19/45


20/45

The hierarchy of causes Multiple causes and risk factors can often be displayed in

the form of a hierarchy of causes, where some are theproximal or most immediate causes (precipitating factors)and others are distal or indirect causes (enabling factors).Inhaled tobacco smoke is a proximal cause of lungcancer, while low socio-economic status is a distal causethat is associated with smoking habits and indirectly withlung cancer. Various frameworks have been devised forvisualizing the relationships between the distal and

proximal causes and the eventual health effects. Onesuch multi-layer framework, DPSEEA (driving forces,pressure, state, exposure, effect, action), was used byWHO to analyze different elements of causation,prevention and indicators in relation to environmentalhealth hazards (Figure 5.5).


21/45


22/45

A similar framework was developed for the WHO GlobalBurden of Disease project. The Multiple Exposures, MultipleEffects framework emphasizes the complex relationships

between environmental exposures and child healthoutcomes. This model takes into account that individualexposures can lead to many different health outcomes, andspecific health outcomes can be attributed to many differentexposures.

In epidemiological studies linking one or more causes to ahealth outcome, it is important to consider to what extentdifferent causes are at the same or different levels in thehierarchy. If a cause of a cause is included in the analysistogether with the cause itself, the statistical method ofanalysis has to take this into account. The identification of

the hierarchy of causes and the quantitative relationshipsbetween them will provide one way of describing themechanism of causation. For example, low socio-economicstatus is associated in many industrialized nations with moretobacco smoking, which is associated with higher bloodpressure, which in turn increases the risk of stroke.


23/45

Establishing the cause of the

disease Causal inference is the term used for the process of

determining whether observed associations are likely to becausal; the use of guidelines and the making of judgments

are involved. The process of judging causation can bedifficult and contentious. It has been argued that causalinference should be restricted to the measurement of aneffect, rather than as a criterion-guided process for decidingwhether an effect is present or not. Before an association is

assessed for the possibility that it is causal, otherexplanations, such as chance, bias and confounding, haveto be excluded. The steps in assessing the nature of therelationship between a possible cause and an outcome areshown in Figure 5.6.


24/45


25/45

Considering Causation A systematic approach to determining the nature

of an association was used by the United StatesSurgeon General to establish that cigarette

smoking caused lung cancer. This approachwas further elaborated by Hill.11 On the basis ofthese concepts, a set of considerations forcausation, listed in the sequence of testing thatthe epidemiologist should follow to reach aconclusion about a cause of disease, is shown inTable 5.1.


26/45


27/45

Temporal Relationship The temporal relationship is crucialthe cause must

precede the effect. This is usually self-evident, althoughdifficulties may arise in case-control and cross-sectional

studies when measurements of the possible cause andeffect are made at the same time. In cases where thecause is an exposure that can be at different levels, it isessential that a high enough level be reached before thedisease occurs for the correct temporal relationship toexist. Repeated measurement of the exposure at morethan one point in time and in different locations maystrengthen the evidence.


28/45

Plausibility An association is plausible, and thus more likely

to be causal, if consistent with other knowledge.

For instance, laboratory experiments may haveshown how exposure to the particular factorcould lead to changes associated with the effectmeasured. However, biological plausibility is arelative concept, and seemingly implausible

associations may eventually be shown to becausal.


29/45

Consistency Consistency is demonstrated by several studies giving the

same result. This is particularly important when a variety ofdesigns are used in different settings, since the likelihood

that all studies are making the same mistake isminimized. However, a lack of consistency does notexclude a causal association, because different exposurelevels and other conditions may reduce the impact of thecausal factor in certain studies. Furthermore, when theresults of several studies are being interpreted, the best-

designed ones should be given the greatest weight.


30/45

Techniques are available for pooling the results ofseveral studies that have examined the same issue,particularly randomized controlled trials.

This technique is called meta-analysis and is used

to combine the results of several trials, each ofwhich may deal with a relatively small sample, toobtain a better overall estimate of effect (Figure 5.7).


31/45


32/45

Strength

A strong association between possible cause and effect,as measured by the size of the risk ratio (relative risk), ismore likely to be causal than is a weak association,

which could be influenced by confounding or bias.Relative risks greater than 2 can be considered strong.For example, cigarette smokers have a twofold increasein the risk of acute myocardial infarction compared withnon-smokers. The risk of lung cancer in smokers,compared with nonsmokers, has been shown in various

studies to be increased between fourfold and twentyfold. However, associations

of such magnitude are rare in epidemiology.


33/45

The fact that an association is weak does not preclude it from being

causal; the strength of an association depends on the relativeprevalence of other possible causes. For example, weak associationshave been found between diet and risk of coronary heart disease inobservational studies; and although experimental studies on selectedpopulations have been done, no conclusive results have beenpublished. Despite this lack of evidence, diet is generally thought to bea major causative factor in the high rate of coronary heart disease in

many industrialized countries. The probable reason for the difficulty in identifying diet as a risk factor

for coronary heart disease is that diets in populations are ratherhomogeneous and variation over time for one individual is greater thanthat between people. If everyone has more or less the same diet, it isnot possible to identify diet as a risk factor. Consequently, ecologicalevidence gains importance. This situation has been characterized as

one of sick individuals and sick population, 23 meaning that in manyhigh-income countries, whole populations are at risk from an adversefactor.


34/45

Dose-response Relationship

A doseresponse relationship occurs whenchanges in the level of a possible cause areassociated with changes in the prevalence or

incidence of the effect. Table 5.2 illustrates thedoseresponse relationship between noise andhearing loss: the prevalence of hearing lossincreases with noise level and exposure time.The demonstration of such a clear dose

response relationship in unbiased studiesprovides strong evidence for a causalrelationship between exposure and disease.


35/45


36/45

Reversibility

When the removal of a possible cause results in areduced disease risk, there is a greater likelihood thatthe association is causal. For example, the cessation of

cigarette smoking is associated with a reduction in therisk of lung cancer relative to that in people who continueto smoke (see Figure 8.5). This finding strengthens thelikelihood that cigarette smoking causes lung cancer. Ifthe cause leads to rapid irreversible changes that

subsequently produce disease whether or not there iscontinued exposure, then reversibility cannot be acondition for causality.


37/45

Study Design

The ability of a study design to provecausation is an important consideration.

Table 5.3 outlines the different types ofstudy and relative strengths in establishingcausality. These study designs wereintroduced in Chapter 3; their use in

providing evidence for causal relationshipsis discussed below.


38/45


39/45

Experimental studies

The best evidence comes from well-designed

randomized controlled trials. However, evidence

is rarely available from this type of study, and

often only relates to the effects of treatment and

prevention campaigns. Other experimental

studies, such as field and community trials, are

seldom used to study causation.


40/45

Cohort studies and case control studies

Cohort studies are the next best designbecause, when well conducted, bias isminimized. Again, they are not always available.Although case-control studies are subject to

several forms of bias, the results from large,well-designed investigations of this kind providegood evidence for the causal nature of anassociation; judgements often have to be madein the absence of data from other sources.


41/45

Cross-sectional studies Cross-sectional studies are less able to prove causation

as they provide no direct evidence on the time sequenceof events. However, the time sequence can often beinferred from the way exposure and effect data is

collected. For instance, if it is clear that the health effectis recent and the exposure to the potential causes isrecorded in a questionnaire, questions about the pastmay clearly identify exposures before the effectoccurred.


42/45

Ecological studies

Ecological studies provide the weakest evidence forcausality because of the danger of incorrectextrapolation to individuals from regional or nationaldata. However, for certain exposures that cannotnormally be measured individually (such as air pollution,pesticide residues in food, fluoride in drinking water),evidence from ecological studies is very important. Whencausal relationships have already been established, well-designed ecological studies, particularly time series

studies, can be very useful to quantify effects.


43/45

Judging the evidence

Regrettably, there are no completely reliable criteria for determiningwhether an association is causal or not. Causal inference is usuallytentative and judgements must be made on the basis of theavailable evidence: uncertainty always remains. Evidence is often

conflicting and due weight must be given to the different types whendecisions are made. In judging the different aspects of causationreferred to above, the correct temporal relationship is essential;once that has been established, the greatest weight may be given toplausibility, consistency and the doseresponse relationship. Thelikelihood of a causal association is heightened when many differenttypes of evidence lead to the same conclusion.


44/45

Evidence from well-designed studies is particularly important,

especially if they are conducted in a variety of locations. The mostimportant use of information about causation of diseases andinjuries may be in the area of prevention, which we will discuss inthe following chapters. When the causal pathways are establishedon the basis of quantitative information from epidemiological studies,the decisions about prevention may be uncontroversial. In situationswhere the causation is not so well established, but the impacts have

great potential public health importance, the precautionaryprinciple29 may be applied to take preventive action as a safetymeasure; this is called precautionary prevention.


45/45

THANK

YOU!!!

epidemiology concept of cause

Documents